A continuous improvement framework for AI agents to log errors, learnings, and corrections into persistent project memory.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install perrytest
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install perrytest using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Self-Improvement Skill?
The Self-Improvement Skill is a specialized toolset designed for AI coding agents to facilitate autonomous growth and refinement. By capturing unexpected failures, user corrections, and discovered best practices, it ensures that an agent does not repeat the same mistakes across different sessions. This skill is a cornerstone of the Openclaw Skills ecosystem, enabling agents to build a robust internal knowledge base that matures alongside the codebase through systematic documentation.
Self-Improvement Skill Use Cases
- Capturing unexpected command or API failures to prevent future friction and repetitive troubleshooting.
- Documenting user corrections when the agent makes a logical, stylistic, or architectural error.
- Tracking feature requests for capabilities that do not yet exist in the current environment.
- Identifying and promoting recurring patterns into permanent project documentation like CLAUDE.md or AGENTS.md.
- Sharing learnings between different AI sessions or sub-agents using OpenClaw workspace tools.
How Self-Improvement Skill Works
- Detection: The agent identifies a learning opportunity, such as a tool error, knowledge gap, or a direct user correction.
- Logging: Detailed entries are appended to specialized markdown files in the .learnings/ directory using a standardized ID format.
- Categorization: Entries are assigned priorities (Low to Critical) and functional areas (Frontend, Backend, etc.) for easier filtering.
- Promotion: High-value, broadly applicable learnings are promoted to long-term memory files like AGENTS.md or SOUL.md.
- Extraction: Proven solutions that are verified and recurring are extracted into standalone, reusable Openclaw Skills.
Self-Improvement Skill Setup
Installing the skill via ClawdHub is the recommended path for Openclaw Skills users:
clawdhub install self-improving-agent
For manual installation in your workspace:
git clone https://github.com/peterskoett/self-improving-agent.git ~/.openclaw/skills/self-improving-agent
Initialize the required directory structure:
mkdir -p ~/.openclaw/workspace/.learnings
Self-Improvement Skill Data Schema & Taxonomy
The skill maintains data within a .learnings/ directory using a structured file system to ensure transparency and readability:
| File | Purpose |
|---|---|
| LEARNINGS.md | Stores corrections, knowledge gaps, and best practices. |
| ERRORS.md | Logs command failures, exceptions, and environmental issues. |
| FEATURE_REQUESTS.md | Tracks capabilities requested by the user. |
Each entry follows a strict ID format: TYPE-YYYYMMDD-XXX (e.g., LRN-20250115-001) and includes metadata like Priority, Area, Status, and Pattern-Key.
Self-Improvement Skill Advanced Features
- Automated Hooks: Support for UserPromptSubmit and PostToolUse hooks to trigger error detection and evaluation reminders automatically.
- Inter-Session Messaging: Built-in tools like sessions_send and sessions_history to share critical learnings across active agent sessions.
- Pattern Deduplication: Seamless ingestion of recurring patterns via the simplify-and-harden feed using stable Pattern-Keys for tracking frequency.
- Skill Extraction Helper: A dedicated automated script to transform a verified learning into a fully-compliant, reusable skill folder with its own metadata.
- Workspace Promotion: Automatic logic for elevating tactical guidance into system-level prompt files when specific recurrence thresholds are met.
SKILL.md
Loading
Related Openclaw Skills
Dependency Vulnerability Scanner
A zero-config security tool that scans project dependencies for known vulnerabilities using the OSV.dev database.
Ecologi
A powerful integration for AI agents to manage carbon offsets, fund tree planting, and track environmental impact statistics via Ecologi.
Goal Setting & Achievement Reference
A comprehensive technical reference for implementing research-backed goal-setting frameworks and habit formation systems.
Hydrate
A command-line reference for managing hydration science, fluid intake calculations, and electrolyte balance.
Clawdit Borrower
An autonomous finance skill that enables agents to borrow USDT from P2P lenders using WDK-powered smart wallets and ERC-8004 reputation profiles.
Copy
A comprehensive reference skill for mastering file duplication, rsync workflows, and efficient data synchronization strategies.
Featured*
